Evolving and Executing Research Plans via Double-Loop Multi-Agent Collaboration

2025-10-08

AI Summary

This work addresses the challenge of achieving end-to-end automation across the entire scientific research lifecycle, where dynamic evolution of high-level research plans and reliable execution under uncertainty pose fundamental obstacles. We propose a novel dual-loop multi-agent collaboration framework: an outer loop, guided by a “Professor” agent, employs evolutionary algorithms to iteratively optimize research strategies; an inner loop, orchestrated by a “PhD Student” agent, enables real-time execution adaptation via context-aware pre- and post-meeting protocols, dynamic task scheduling, and environmental observation. Crucially, this framework formalizes plan evolution and execution feedback as a closed-loop co-adaptive process—its first such instantiation. Evaluated on benchmarks including ACLAward and Laboratory, our system achieves state-of-the-art scores in automatically generated paper quality, significantly outperforming strong baselines. Ablation studies confirm the critical contribution of the dual-loop design to overall performance gains.

Technology Category

• Multiagent Systems: Multiagent Planning
• Humans and AI: Planning and Decision Support for Human-Machine Teams
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Research challenges in human and human-AI computation

Abstract

Automating the end-to-end scientific research process poses a fundamental challenge: it requires both evolving high-level plans that are novel and sound, and executing these plans correctly amidst dynamic and uncertain conditions. To address this bilevel challenge, we propose a novel Double-Loop Multi-Agent (DLMA) framework to solve the given research problem automatically. The leader loop, composed of professor agents, is responsible for evolving research plans. It employs an evolutionary algorithm through involvement, improvement, and integration meetings to iteratively generate and refine a pool of research proposals, exploring the solution space effectively. The follower loop, composed of doctoral student agents, is responsible for executing the best-evolved plan. It dynamically adjusts the plan during implementation via pre-hoc and post-hoc meetings, ensuring each step (e.g., drafting, coding) is well-supported by contextual and external observations. Extensive experiments on benchmarks like ACLAward and Laboratory show that DLMA generates research papers that achieve state-of-the-art scores in automated evaluation, significantly outperforming strong baselines. Ablation studies confirm the critical roles of both loops, with evolution driving novelty and execution ensuring soundness.

Problem

• Automating end-to-end scientific research with evolving plans
• Executing research plans under dynamic uncertain conditions
• Generating novel sound research through multi-agent collaboration

Innovation

• Double-loop multi-agent framework automates research
• Evolutionary algorithm refines research proposals iteratively
• Dynamic plan adjustment ensures robust execution steps